Data-driven methods such as statistical process control and machine learning often struggle to capture complex, non-linear relationships or lack interpretability, thus hindering the overall adoption success story of anomaly detection in complex industrial processes. Anomalies can manifest in diverse and often subtle ways and may not be directly evident from readily observable metrics. This work proposes a novel approach that leverages causal discovery to identify critical process parameters and their underlying causal relationships. By focusing on causal mechanisms, this approach aims to reduce the complexity and enhance the interpretability and robustness of anomaly detection in the industrial context. The proposed methodology is applied to a real-world blast furnace plant, where it demonstrated the potential to uncover interdependencies, detect anomalies, their key drivers, and provide insights into the underlying mechanisms of the process. The results show that modeling based on the causal information decreases the complexity of the model-based anomaly detection, validating the applicability of this approach in other fields of science and engineering.

Abstract The detection, description and understanding of anomalies in multivariate time series data is an important task in several industrial domains. Automated data analysis provides many tools and algorithms to detect anomalies, while visual interfaces enable domain experts to explore and analyze data interactively to gain insights using their expertise. Anomalies in multivariate time series can be diverse with respect to the dimensions, temporal occurrence and length within a dataset. Their detection and description depend on the analyst's domain, task and background knowledge. Therefore, anomaly analysis is often an underspecified problem. We propose a visual analytics tool called MANDALA (Multivariate ANomaly Detection And expLorAtion), which uses kernel density estimation to detect anomalies and provides users with visual means to explore and explain them. To assess our algorithm's effectiveness, we evaluate its ability to identify different types of anomalies using a synthetic dataset generated with the GutenTAG anomaly and time series generator. Our approach allows users to define normal data interactively first. Next, they can explore anomaly candidates, their related dimensions and their temporal scope. Our carefully designed visual analytics components include a tailored scatterplot matrix with semantic zooming features that visualize normal data through hexagonal binning plots and overlay candidate anomaly data as scatterplots. In addition, the system supports the analysis on a broader scope involving all dimensions simultaneously or on a smaller scope involving dimension pairs only. We define a taxonomy of important types of anomaly patterns, which can guide the interactive analysis process. The effectiveness of our system is demonstrated through a use case scenario on industrial data conducted with domain experts from the automotive domain and a user study utilizing a public dataset from the aviation domain.

Visual analysis (VA) tasks often involve exploring large and complex multi-dimensional datasets to identify trends and anomalies. However, the challenge lies in displaying all the data and maintaining the desired level of detail within the limited screen space. In this paper, we propose a solution that incorporates multiple visualizations and semantic zooming to address this compromise. Our visualization tool focuses on cycle-dependent data, showcasing time series with repetitive behavior. Through semantic zooming, cyclic time series data can be displayed in large quantities and high levels of detail without the need for multiple views. Our proposed tool includes three independent visualizations: line plots, horizon graphs, and adaptive heatmaps. By offering different visualization options, we aim to provide a rich and flexible analytical experience that response to the different user needs and encourages comprehensive data exploration. The tool accommodates both novice and expert users, allowing for intuitive analysis as well as advanced techniques for detailed examination. Our approach follows the mantra of “overview first, zoom and filter, then details-on-demand” facilitating rapid detection and exploration of patterns and trends. In this paper, we present the detailed design, interaction capabilities with semantic zoom, and the results of a user study that demonstrate the effectiveness and usefulness of our proposed tool.

Recovering causality from copious time series data beyond mere correlations has been an important contributing factor in numerous scientific fields. Most existing works assume linearity in the data that may not comply with many real-world scenarios. Moreover, it is usually not sufficient to solely infer the causal relationships. Identifying the correct time delay of cause-effect is extremely vital for further insight and effective policies in inter-disciplinary domains. To bridge this gap, we propose KOMPOS, a novel algorithmic framework that combines a powerful concept from causal discovery of additive noise models with graphical ones. We primarily build our structural causal model from multivariate adaptive regression splines with inherent additive local nonlinearities, which render the underlying causal structure more easily identifiable. In contrast to other methods, our approach is not restricted to Gaussian or non-Gaussian noise due to the non-parametric attribute of the regression method. We conduct extensive experiments on both synthetic and real-world datasets, demonstrating the superiority of the proposed algorithm over existing causal discovery methods, especially for the challenging cases of autocorrelated and non-stationary time series.

Visualizations have a distinctive advantage when dealing with the information overload problem: Because they are grounded in basic visual cognition, many people understand them. However, creating proper visualizations requires specific expertise of the domain and underlying data. Our quest in this article is to study methods to suggest appropriate visualizations autonomously. To be appropriate, a visualization has to follow known guidelines to find and distinguish patterns visually and encode data therein. A visualization tells a story of the underlying data; yet, to be appropriate, it has to clearly represent those aspects of the data the viewer is interested in. Which aspects of a visualization are important to the viewer? Can we capture and use those aspects to recommend visualizations? This article investigates strategies to recommend visualizations considering different aspects of user preferences. A multi-dimensional scale is used to estimate aspects of quality for visualizations for collaborative filtering. Alternatively, tag vectors describing visualizations are used to recommend potentially interesting visualizations based on content. Finally, a hybrid approach combines information on what a visualization is about (tags) and how good it is (ratings). We present the design principles behind VizRec , our visual recommender. We describe its architecture, the data acquisition approach with a crowd sourced study, and the analysis of strategies for visualization recommendation.

Recommender systems are becoming common tools supporting automatic, context-based retrieval of resources. When the number of retrieved resources grows large visual tools are required that leverage the capacity of human vision to analyse large amounts of information. We introduce a Web-based visual tool for exploring and organising recommendations retrieved from multiple sources along dimensions relevant to cultural heritage and educational context. Our tool provides several views supporting filtering in the result set and integrates a bookmarking system for organising relevant resources into topic collections. Building upon these features we envision a system which derives user's interests from performed actions and uses this information to support the recommendation process. We also report on results of the performed usability evaluation and derive directions for further development.